13. Data Flashcards
Sources of data
- own data
- population data (provided by government)
- reinsurer data
- market data (insured lives data, published returns)
- other sources (trade magazines, actuarial consultants, overseas data, rate table software)
Desirable characteristics of data
- relevant (relate to similar class of lives and policy conditions to those expected)
- credible (sufficiently large volume of data)
- available
- reliable for intended purposes
Data required to estimate claim rate parameters for stand-alone CI insurance
- Date of birth, entry, diagnosis. notification of sickness, claim payment
- reason for exit, type of disease
The method is to construct central exposures to risk, grouped by age, for healthy policyholders - number of claims due to each particular disease/ procedure arising from exposure, grouped by age, should also be counted
- claim inception can be estimated = number claims/ exposure
Data required to estimate claim rate parameters for LTCI
- dates of birth, entry, exit, diagnosis, first becoming incapacitated, notification of incapacity, start of LTC claim, end of LTC claim
- reasons for exit/ end of LTC claim
- construct central exposure to risk, grouped by age, separately for claiming and healthy policyholders
- number of transitions arising from exposure also needs to be counted
- require dates of transition from one level to another
PMI
- count all policies taken out/ renewed during year of investigation
- count number of claim incidences arising from each year’s cohorts of policies, subdivided by age and class of benefit / procedure involved
- record total amount of claims paid from each cohort
In practice
- calculate number policy years of exposure during a chosen investigation period
- count number of claim incidences arising from this exposure, subdivided by age and class of benefit
- record the total amount of claims paid for this accident year
rates should be projected to mid-point of insurance usage
Own data
- best base for prediction
- cost, reliability and format should not be a problem
- credibility lower in health and care, insufficient data, greater reliance on other sources
Population data
- govt produced periodic analysis of healthcare experience of population
- accuracy and reliability may be questionable - definition subjective and double counting significant
- may be out-dated if population
- may not be fully relevant to insured lives
- definition of sickness - not equivalent to definition of sickness that triggers claim
- readily available, free, large volume and thus credible
Reinsurer data
- accurate, as profits under treaty will reflect premiums calculated
- reinsurer should make date relevant
- adjustments required to allow for company specific underwriting and claims management practices
- should be more credible than own data
- particularly in context of new product launch, reinsurer likely to provide data in return for share of business
Market data (industry data, published returns)
industry data represents insured experience, generally sufficient volume to be credible, reflective of market practices and compiled by experts
- remains market average and hence not strictly relevant to one company
- by time it is published, it may be out-dated
- more credible than own data, less relevant
- less credible than population data, more relevant and relates to policy claim experience
Returns
- accurate, hard to extract useful information
- used as overall check
Other
overseas - may not be relevant, potentially credible
- actuarial consultant - similar to reinsurer data, consultants likely to have less interest in risk
- rate table software - useful in checking final premium rates
- trade magazines - useful in checking final results
Other
overseas - may not be relevant, potentially credible
- actuarial consultant - similar to reinsurer data, consultants likely to have less interest in risk
- rate table software - useful in checking final premium rates
- trade magazines - useful in checking final results